u n i ve r s i t y o f co pe n h ag e n

Københavns Universitet

Convergent Akt activation drives acquired EGFR inhibitor resistance in lung cancer

Jacobsen, Kirstine; Bertran-Alamillo, Jordi; Molina, Miguel Angel; Teixidó, Cristina;Karachaliou, Niki; Pedersen, Martin Haar; Castellví, Josep; Garzón, Mónica; Codony-Servat,Carles; Codony-Servat, Jordi; Giménez-Capitán, Ana; Drozdowskyj, Ana; Viteri, Santiago;Larsen, Martin R; Lassen, Ulrik; Felip, Enriqueta; Bivona, Trever G; Ditzel, Henrik J; Rosell,RafaelPublished in:Nature Communications

DOI:10.1038/s41467-017-00450-6

Publication date:2017

Document versionPublisher's PDF, also known as Version of record

Document license:CC BY

Citation for published version (APA):Jacobsen, K., Bertran-Alamillo, J., Molina, M. A., Teixidó, C., Karachaliou, N., Pedersen, M. H., ... Rosell, R.(2017). Convergent Akt activation drives acquired EGFR inhibitor resistance in lung cancer. NatureCommunications, 8, [410]. https://doi.org/10.1038/s41467-017-00450-6

Download date: 09. feb.. 2020

ARTICLE

Convergent Akt activation drives acquired EGFRinhibitor resistance in lung cancerKirstine Jacobsen1, Jordi Bertran-Alamillo2, Miguel Angel Molina2, Cristina Teixidó2, Niki Karachaliou3,

Martin Haar Pedersen1, Josep Castellví2, Mónica Garzón2, Carles Codony-Servat2, Jordi Codony-Servat2,

Ana Giménez-Capitán2, Ana Drozdowskyj4, Santiago Viteri5, Martin R. Larsen6, Ulrik Lassen7, Enriqueta Felip8,

Trever G. Bivona9,10, Henrik J. Ditzel 1,11 & Rafael Rosell2,5,12,13

Non-small-cell lung cancer patients with activating epidermal growth factor receptor (EGFR)

mutations typically benefit from EGFR tyrosine kinase inhibitor treatment. However, virtually

all patients succumb to acquired EGFR tyrosine kinase inhibitor resistance that occurs via

diverse mechanisms. The diversity and unpredictability of EGFR tyrosine kinase inhibitor

resistance mechanisms presents a challenge for developing new treatments to overcome

EGFR tyrosine kinase inhibitor resistance. Here, we show that Akt activation is a convergent

feature of acquired EGFR tyrosine kinase inhibitor resistance, across a spectrum of diverse,

established upstream resistance mechanisms. Combined treatment with an EGFR tyrosine

kinase inhibitor and Akt inhibitor causes apoptosis and synergistic growth inhibition in

multiple EGFR tyrosine kinase inhibitor-resistant non-small-cell lung cancer models. More-

over, phospho-Akt levels are increased in most clinical specimens obtained from EGFR-

mutant non-small-cell lung cancer patients with acquired EGFR tyrosine kinase inhibitor

resistance. Our findings provide a rationale for clinical trials testing Akt and EGFR inhibitor

co-treatment in patients with elevated phospho-Akt levels to therapeutically combat the

heterogeneity of EGFR tyrosine kinase inhibitor resistance mechanisms.

DOI: 10.1038/s41467-017-00450-6 OPEN

1 Department of Cancer and Inflammation Research, Institute of Molecular Medicine, University of Southern Denmark, 5000 Odense, Denmark. 2 Laboratoryof Oncology, Pangaea Biotech, Quiron Dexeus University Hospital, 08028 Barcelona, Spain. 3 Instituto Oncológico Dr. Rosell, University Hospital Sagrat Cor,08029 Barcelona, Spain. 4 Pivotal, 28023 Madrid, Spain. 5 Instituto Oncológico Dr. Rosell, Quiron-Dexeus University Hospital, 08028 Barcelona, Spain.6 Protein Research Group, Institute of Biochemistry and Molecular Biology, University of Southern Denmark, 5230 Odense, Denmark. 7 Phase I Unit,Rigshospitalet, 2100 Copenhagen, Denmark. 8Department of Medical Oncology, Vall D´Hebron, 08035 Barcelona, Spain. 9 Department of Medicine, Divisionof Hematology and Oncology, University of California, San Francisco, CA 94158, USA. 10 Helen Diller Comprehensive Cancer Center, University of California,San Francisco, CA 94158, USA. 11 Department of Oncology, Odense University Hospital, 5000 Odense, Denmark. 12 Catalan Institute of Oncology, HospitalGermans Trias i Pujol, 08916 Badalona, Spain. 13 Germans Trias i Pujol, Health Sciences Institute and Hospital, Campus Can Ruti, 08916 Badalona, Spain.Correspondence and requests for materials should be addressed to T.G.B. (email: trever.bivona@ucsf.edu) or to H.J.D. (email: hditzel@health.sdu.dk)

NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications 1

Lung cancer is the leading cause of cancer mortalityworldwide1. Mutations in epidermal growth factor receptor(EGFR), most commonly deletions in exon 19 (delE746-750)

or substitution of arginine for leucine (L858R) in exon 21, arepresent in ~17% of tumors in patients with pulmonary adeno-carcinoma2 and confer sensitivity to the EGFR-tyrosine kinaseinhibitors (TKIs) gefitinib3, 4, erlotinib5, 6 or afatinib7, 8. Theprincipal downstream pathways mediating the oncogenic effectsof EGFR are extracellular signal–regulated kinase 1 and 2 (ERK1/2) via Ras, Akt via phosphatidylinositol 3-kinase (PI3K), and

signal transducer and activator of transcription 3 (STAT3) viaJanus kinase 2 (JAK2)9.

Acquired resistance substantially limits the clinical efficacy ofEGFR TKIs. Although ~70% of EGFR-mutant non-small-celllung cancer (NSCLC) patients respond to first-line EGFR-TKItreatment, the majority of them do not achieve completeresponses and virtually all patients develop acquired resistanceand lethal disease progression6. A diversity of EGFR-TKIresistance mechanisms has been described, of which the mostfrequent mechanism of resistance to EGFR-TKI treatment is the

AXL

N-cadherin

E-cadherin

Vimentin

AXL

0

5

10

15

20*

***

MER

MET

0

2

4

6

* * *

FGFR1*

PC9 GR1 GR2 GR3 GR4 GR5 PC9ER

pEphA2

PC

9

PC

9-E

R

PC

9 G

R1

*

PC

9 G

R2

PC

9 G

R3

PC

9 G

R4

*

PC

9 G

R5

– + + + + + +

* T790M

MW (kDa)

250

250

130

130

130

130

250

130

130

35

100

35

25

15

35

EGFR TKI

EGFR

pEGFR (Y1068)

AXL

MET

pMET

FGFR1

MER

EphA2

Vimentin

E-cadherin

Bcl-XL

Bcl-2

BIM

Actin

Rel

ativ

em

RN

A e

xpre

ssio

n

Rel

ativ

em

RN

A e

xpre

ssio

n

Rel

ativ

em

RN

A e

xpre

ssio

n

Rel

ativ

em

RN

A e

xpre

ssio

n

0.0

1.0

0.8

0.6

0.4

0.2

0.0

0.1

0.2

0.3

0.4Resistant to EGFRi

β-catenin

GR5GR4

GR3GR2

GR1

PC9-ER

PC9

GR5GR4

GR3GR2

GR1

PC9-ER

PC9

GR5GR4

GR3GR2

GR1

PC9-ER

PC9

GR5GR4

GR3GR2

GR1

PC9-ER

PC9

a

b c

Fig. 1 Expression of proteins related to acquired resistance in parental and resistant PC9 cell lines. a Immunohistochemistry of selected EMT markers. Scalebar, 100 µm. b Relative mRNA expression of selected genes. Experiments were performed in triplicates and results are shown as mean± SD. Asterisksindicate significant difference from PC9 in a Student’s t-test. c Western blot analysis of selected proteins. In all cases, PC9 cells were grown without drug,whereas PC9-GR1-5 were grown in 5 µM gefitinib and PC9-ER was grown in 30 µM erlotinib

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6

2 NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications

secondary mutation in exon 20 of EGFR, T790M10, 11. Othermechanisms include amplification, overexpression, and autocrineloops involving MET proto-oncogene (MET), erb-b2 receptortyrosine kinase 2 (ErbB2), ephrin type-A receptor 2 (EphA2),fibroblast growth factor receptor (FGFR) and the members ofthe TAM receptor tyrosine kinase (RTKs), Mer and AXL12–15.In addition, we have shown that activation of NF-κB rescuesEGFR-mutant lung cancer cells from EGFR-TKI treatment16.Finally, BRAF and PIK3CA mutations, conversion to small-cell-lung cancer and occurrence of epithelial-to-mesenchymaltransition (EMT) have also been associated with acquiredresistance to EGFR-TKI in NSCLC12. Certain EGFR-mutantNSCLCs harbor multiple mechanisms of EGFR-TKI resistance17, 18.In these cases, the co-occurrence of multiple resistance mechan-isms is likely to lessen the therapeutic impact of targeting eachindividual resistance-promoting alteration. Additionally, whichspecific resistance alteration(s) will arise and promote EGFR-TKIresistance in individual patients is currently largely unpredictableat the outset of therapy. Hence, the diversity and unpredictabilityof EGFR-TKI resistance mechanisms presents a major challengefor efficiently developing new treatment regimens that canovercome EGFR-TKI resistance in patients.

Activation of the Akt pathway is a common feature inhuman cancers and leads to increased cell survival, growth, andproliferation19. V-akt murine thymoma viral oncogene homologs1, 2 and 3 (Akt1, Akt2, and Akt3) comprise the Akt family ofserine-threonine kinases, which are tethered to the membrane viainteraction with phosphatidylinositol-3,4,5-triphosphate (PIP3)lipids20, and activated by phosphorylation on threonine 308(Thr308) by 3-phosphoinositide-dependent protein kinase 1(PDK1)21 and serine 473 (Ser473) by the mammalian targetof rapamycin complex 2 (mTORC2)22. Activated Aktphosphorylates many downstream targets, including forkheadbox O3 (FOXO3) and proline-rich Akt substrate of 40 kDa(PRAS40)23–26. Several small molecule drugs targetingcomponents of the Akt pathway have been developed and arebeing tested in patients27. Interestingly, first-line sensitivity toEGFR TKIs in NSCLC has been associated with pre-existent Aktactivation that is suppressed by EGFR inhibition, while treatmentwith EGFR TKIs failed to block Akt signaling in tumor cellsintrinsically resistant to these drugs28–31. In addition, thecombination of a PIK3-mTOR inhibitor with a MEK inhibitorhas been reported to induce apoptosis in EGFR-TKI naïveEGFR-mutant NSCLC cell lines and xenografts, although thecombination of an Akt and a MEK inhibitor failed to have thiseffect in this TKI-naive context32. Despite evidence suggestinga general role for PI3K-AKT-mTOR pathway signaling inEGFR-mutant NSCLC, whether Akt activation, specifically, candrive acquired EGFR-TKI resistance has not been clearlydemonstrated. Furthermore, the hypothesis that Akt activationfunctions as a convergent, resistance-driving signalingevent across a spectrum of EGFR-mutant NSCLCs that harborotherwise diverse, established EGFR-TKI resistance-promotingmechanisms has not been tested.

Here, we show that Akt pathway activation is a convergentfeature in EGFR-mutant NSCLCs with acquired resistanceto EGFR TKIs that may be caused by diverse underlyingmechanisms. This convergent resistance-promoting function ofAkt activation occurred in the presence of one or moredifferent resistance mechanisms such as amplification,overexpression, and activation of MET, EphA2, FGFR, Mer, andAXL or the presence of the T790M mutation. We show thatcombined treatment with Akt and EGFR inhibitors in resistantEGFR-mutant NSCLC models synergistically inhibits growth inthis heterogeneous molecular background. We also show thatphospho-Akt (pAkt) is increased in the majority of EGFR-mutant

patients after progression on EGFR TKIs, and also that high levelsof pAkt in patients prior to EGFR-TKI treatment correlates withpoor initial therapeutic responses. Thus, convergent Akt activa-tion is a previously unrecognized therapeutic target whose inhi-bition has the unique potential to combat the profound molecularheterogeneity underlying EGFR-TKI clinical resistance. Ourfindings provide a new rationale to test combined Akt plus EGFRinhibitor treatment, specifically in EGFR-mutant NSCLCs withhigh pAkt levels, to overcome acquired EGFR-TKI resistance andenhance patient survival.

ResultsEstablishment of PC9-derived cell lines resistant to EGFRTKIs. Six EGFR-TKI-resistant cell lines were generated bytreating EGFR-TKI-sensitive PC9 cells, which harbor the EGFRexon 19 deletion, with increasing concentrations of gefitinib(GR1-5) or erlotinib (ER). Sequencing analyses revealed that allsix cell lines retained the EGFR exon 19 deletion (SupplementaryTable 1), while the T790M mutation emerged in two (PC9-GR1and PC9-GR4 at allelic fractions of 25 and 38% respectively).No mutations were detected in HER2, PIK3CA, BRAF of KRAS.We determined the sensitivity of the parental and resistant celllines to a variety of EGFR TKIs (Supplementary Table 2). Thehalf-maximal inhibitory concentration (IC50) for gefitiniband erlotinib of parental PC9 cells was in the nanomolarrange compared to 4–29 µM in the resistant cell lines. In T790M-negative cells (PC9-ER, -GR2, -GR3, and -GR5), acquisitionof resistance to EGFR-TKI also led to insensitivity to the second-(afatinib or dacomitinib) and third- (osimertinib (AZD9291))generation EGFR TKIs. The T790M-positive cell lines (PC9-GR1and -GR4) remained sensitive to osimertinib and became onlymoderately resistant to afatinib and dacomitinib. Acquisition ofthe resistant phenotype was often associated with morphologicalchanges and altered anchorage-independent growth, invasiveness,and migration (Supplementary Fig. 1).

Common resistance mechanisms in PC9-derived cell lines.To delineate the underlying resistance mechanism of theEGFR-TKI-resistant cell lines, we initially investigated previouslyreported acquired resistance mechanisms other than T790Mmutation using fluorescence in situ hybridization, immunohis-tochemistry (IHC), quantitative real-time polymerase chainreaction (qRT-PCR) and wsestern blotting. Regarding mRNA andprotein expression, AXL upregulation was the most commonmolecular event, while FGFR1 overexpression and EphA2 andMET activation were also found in some of the resistant celllines (Figs. 1a–c). In contrast, no significant differences wereobserved in the case of FGFR2, MER (Figs. 1b, c), ErbB2, orErbB3 (Supplementary Fig. 2). Alterations in proteins related toapoptosis (Bcl-2, Bcl-xl, and BIM) and EMT (E-cadherin,vimentin, N-cadherin, and beta-catenin) were also observed(Figs. 1a, c). EGFR was equally amplified in the parental PC9 andall the resistant cell lines, while there was no MET, FGFR1, orErbB2 amplification in any of them. In all six resistant cell lines,two or more possible mechanisms of acquired resistance weresimultaneously present (Supplementary Table 3), but each cellline had a unique profile in terms of the specific repertoire ofestablished resistance mechanisms.

Quantitative proteomic analysis of PC9-ER resistant cells.To further identify mechanisms associated with EGFR-TKIresistance, we used quantitative proteomics to compare theproteomes of parental and PC9-ER cells. To increase the numberof proteins identified, samples were first separated into solubleand membrane-associated proteins and subsequently fractionated

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6 ARTICLE

NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications 3

by hydrophilic interaction liquid chromatography (HILIC). Eachof the ten HILIC fractions was subsequently analyzed by liquidchromatography tandem mass spectrometry (LC–MS/MS). PC9and PC9-ER cells were analyzed in biological triplicates. A total of3535 proteins were identified based on two unique peptides (FDR1%) with an overlap of 63.5% between the membrane-associatedand soluble fractions. Based on the distribution of the log2transformed ratios of differentially expressed proteins, 2-fold waschosen as the differential expression threshold. A total of 357proteins exhibited altered expression, of which 247 wereupregulated and 110 downregulated, in the EGFR-TKI-resistantPC9-ER cells compared to the EGFR-TKI-sensitive, parentalPC9 cells.

Next, the differentially expressed proteins between EGFR-TKI-sensitive and -resistant cells were classified according to functional

subgroups (Supplementary Fig. 3a–c). The extensive number ofproteins differentially expressed and the variety of cell processesaffected revealed an entire reprogramming of the cell machineryassociated with the emergence of resistance. Four membranereceptors were upregulated in the resistant cell lines: AXL,integrin alpha-V (ITGAV) and, to a lesser extent, insulin-likegrowth factor 2 receptor and EGFR. The result for AXL wascoincident with the upregulation shown by western blotting, IHCand qRT-PCR. Regarding intracellular signal transduction, themost relevant finding was the 7-fold upregulation of themammalian target of rapamycin (mTOR). We also observed asignificant overexpression of key proteins such as v-crk aviansarcoma virus CT10 oncogene homolog-like (CRKL), ERK1,ERK2, several G-proteins and GTPase activators, some phospha-tases and the ubiquitin E3 ligase Midline 1 (MID1; Fig. 2a).

a

PC

9

PC

9-E

R

PC

9-G

R1

*

PC

9-G

R2

PC

9-G

R3

PC

9-G

R4

*

PC

9-G

R5

EGFR TKI –

Src

pSrc

AKT

pAKT S473

PTEN

Actin

ERK1/2

pERK1/2

b

–4 –2 0 2 4 6

–2 0 2 4 6

CD44FN1LGALS3BP

CDH1JUP

MYH14EPPK1

SPECC1LLOXL2TAGLNARVCF

0 1 2

CDC37CHORDC1TOMM34

HSPBP1HSPA5

DNAJB4MID1

–2 –1 –2 –10 1 2 3 4

TNIKMAPK3PTPRFPTPRK

EGFRAXL

MAPK1MTOR

PTPRG

–4 –2 0 2 4

Soluble

Membrane associated

+ + + + + +

pSTAT3

STAT3

* T790M

FOXO3a

pFOXO1/3a

PRAS40

pPRAS40

MW (kDa)

35

55

35

35

35

70

70

70

70

55

55

55

55

35

KLHL22BID

CDKN2AKIF4A

APOBEC3BPEA15POLD1

NED9TACC3CD44UBE2CALDH3A1CDC20ALDH1A3

IQGAP3CRKL

CTNND1CYR61

ARHGEF2IGF2R

CTNNB1TACSTD2

CD109GNB1

JUPTRPM4ITGAV

MVPS100P

Signal transductionothersCell cycle and apoptosis

PC9-ER vs. PC9PC9-ER vs. PC9

Resistant to EGFRi

PC9-ER vs. PC9 PC9-ER vs .PC9 PC9-ER vs. PC9

Protein modification andheat shock proteins

Signal transductionKinases and phosphatases

Cytoskeleton

HSP90AB1

Fig. 2 Proteomics and western blot analyses of signal transduction proteins in the parental and EGFR-TKI-resistant PC9 cells. a Distribution of ratiosbetween differentially expressed proteins of the Akt pathway in PC9-ER compared to PC9 in the membrane-associated and soluble fractions derived fromthe mass spectrometric analysis (data are log2 transformed). Ingenuity Pathway Analysis software (Qiagen) was used for classifying the proteins indifferent functional groups. b Western blot analysis of key signal transduction proteins in the parental PC9 cells growing in absence of drugs and the sixEGFR-TKI-resistant PC9-derived cell lines cultured with EGFR TKI

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6

4 NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications

The roles of mTOR, ERK1 and ERK2 are widely known, whileCRKL is an oncogene that constitutes a major convergence pointin tyrosine kinase signaling. In addition, CRKL amplification has

been described as a mechanism of acquired resistance to EGFRTKIs in some EGFR-mutant NSCLCs, acting via activation of theERK and Akt pathways33. MID1 is known to activate the Akt/

b

48 h96 h

PC9-ER PC9-GR1

PC9-GR4 PC9-GR5PC9-GR3PC9-GR2

PC9-ER0

1

2

3

4

PC9-GR10

5

10

15

20

PC9-GR20

1

2

3

4

5

PC9-GR30

2

4

6

PC9-GR40

5

10

15

20

PC9-GR50

5

10

15

No drugEGFRiAKTi (G)AKTi (A)EGFRi + AKTi (G)EGFRi + AKTi (A)

PC9-ER

2 4 6 8 2 4 6 8

2 4 6 8

** *

*

**

**

**

* ** *

*

** *

* **

*

** *

*

** *

*

** *

*

** *

*

** *

*

** *

*

PC90

2

4

6

8

10

2 4 6 8

PC9

** *

*

–0.5

0.0

0.5

1.0

1.5

–0.5

0.0

0.5

1.0

1.5

–0.5

0.0

0.5

1.0

1.5

2.0

–0.5

0.0

0.5

1.0

1.5

2.5

2.0

–0.5

0.0

0.5

1.0

1.5

2.5

2.0

Rel

ativ

e to

t=

0

Rel

ativ

e to

t=

0

Rel

ativ

e to

t=

0

Rel

ativ

e to

t=

0

Rel

ativ

e to

t=

0

Rel

ativ

e to

t=

0

Rel

ativ

e to

t=0

3.0

2.0

1.0

0.0

–1.0

0.8

0.6

0.4

0.2

0.0

–0.2

Rel

ativ

e to

unt

reat

ed %

250

200

150

100

50

0

250

200

150

100

50

0

200

150

100

50

0

200

150

100

50

0

150

100

50

0

150

100

50

0

250

200

150

100

50

0

Apo

ptos

is

Time (days) Time (days) Time (days)

2 4 6 8 2 4 6 82 4 6 8Time (days) Time (days) Time (days) Time (days)

PC9-GR5

PC9-GR1

PC9-GR4PC9-GR3

PC9

PC9-GR2

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

a

c

Fig. 3 Synergistic effects of EGFR TKIs combined with Akt inhibitors in parental and EGFR-TKI-resistant PC9 cells. The combined effects of EGFR inhibitors(erlotinib for PC9-ER and gefitinib for parental PC9 and PC9-GR1-5) and Akt inhibitors (GSK2141795 or AZD5363 for all cell lines) were assessed by acrystal violet viability assay performed over 8 days, b BrdU incorporation assay performed after 48 and 96 h incubation, and c apoptosis assay performedafter 96 h incubation. Concentrations of drugs were chosen to be sub-inhibitory to explore the synergistic potential: erlotinib 30 µM; gefitinib 5 µM (exceptfor PC9 where 40 nM was used); GSK2141795 (G) 2.5 µM; AZD5363 (A) 35 µM, (except for GR2 and GR5 where 1.25 and 3 µM were employed). Allexperiments were conducted in quadruplicates (a), or triplicates (b, c), and results are shown as mean± SD. Asterisks indicate significant difference inANOVA one-way test (p< 0.05) for the drug combination-treated cells compared to cells treated with either drug alone at the same time point

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6 ARTICLE

NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications 5

mTOR pathway through degradation of the catalytic subunit ofthe tumor suppressor phosphatase 2A (PP2A)34.

We subsequently performed a global analysis of the selectedproteins using the Ingenuity Pathway Analysis software.Regarding signal transduction, the most relevant finding wasthe upregulation of the PI3K-Akt-mTOR signaling pathway inthe EGFR-TKI-resistant vs. sensitive cells. This finding wasin agreement with the upregulation of several proteins involved inor activated by the PI3K-Akt-mTOR signaling pathway revealedby the individual analyses, such as the membrane receptors AXL,ITGAV, and IGFR2, the downstream signaling proteins CRKLand mTOR, the ubiquitin ligase MID1 or the lysine oxidaseLOXL2 (Fig. 2a).

Activation of the Akt pathway in PC9-derived resistant celllines. In view of the proteomics results, we explored the activationstatus of Akt signal transduction pathways in the parental andresistant cell lines (Fig. 2b) and also studied ERK and STAT3activation. The levels of pAkt and its downstream effectorphospho-PRAS40 (pPRAS40) were elevated in the presence ofEGFR TKI in all the resistant cell lines, regardless of T790M,AXL, MET, EphA2, or FGFR1 status, while significant levelspFOXO1/3a were also present, indicating that Akt is a convergenthub for a variety of acquired resistance mechanisms. In contrast,key effector proteins of ERK and STAT3 signaling pathwaysdid not show the same consistent behavior (Fig. 2b). The tworesistant T790M-positive cell lines had increased levels ofphospho-ERK (pERK) in the presence of EGFR-TKI,while ERK activation was strongly inhibited by the drug inT790M-negative cell lines. Finally, phosphorylated levels ofSTAT3 and c-Src were only elevated in one and three resistantcell lines, respectively.

To gain insight in the mechanisms leading to Akt activation,resistant cells were treated with selected targeted agents(Supplementary Fig. 4). The FGFR inhibitor nintedanib and theMET inhibitor crizotinib induced a dose-dependent decrease ofpAkt levels in PC9-GR5 and PC9-GR2, respectively. PC9-GR5cells overexpress FGFR1 and PC9-GR2 cells show MET activation(Fig. 1b). In contrast, the AXL inhibitor BGB324 had little orno effect on Akt phosphorylation when added to the

AXL-overexpressing PC9-ER cells. Finally, the T790M-positivePC9-GR1 cells, which also show MET activation, weretreated with crizotinib and osimertinib. Both drugs induced amoderate, dose-dependent decrease in pAKT levels (Supplemen-tary Fig. 4d).

EGFR TKI and an Akt inhibitor elicits synergistic growthinhibition. Of all the RTKs and signal transduction proteinsinvestigated, only Akt and PRAS40 were consistently activatedin the six resistant cell lines in the presence of EGFR TKI.Consequently, we assessed the effects of two different Aktinhibitors, GSK2141795 and AZD5363, on cell growth andsignaling. When tested as single agents, both inhibited cell growthwith IC50s of 5–15 µM (Supplementary Fig. 5a). When combinedwith an EGFR TKI, the Akt inhibitors elicited a synergisticgrowth inhibitory effect on the six resistant cell lines, whichwas significantly less strong in the parental PC9 (Fig. 3a, b andSupplementary Table 4). This synergistic effect was observed atall incubation times tested and with a variety of methods:including crystal violet, measuring the number of cells (Fig. 3a),CellTiterBlue, estimating the metabolic activity of living cells(Supplementary Fig. 5b) and BrdU incorporation, measuring cellproliferation (Fig. 3b). Finally, addition of an Akt inhibitor to theEGFR TKI also had a significant enhancing effect on apoptosis asmeasured by a DNA fragmentation assay (Fig. 3c). In view ofthese results, we examined the effects of the Akt inhibitors andEGFR-TKI treatment on signal transduction pathways in threecell lines: PC9, PC9-ER (T790M-negative) and PC9-GR1(T790M-positive; Fig. 4a–c). EGFR TKIs alone suppressedphospho-EGFR (pEGFR), and pERK to a lesser extent, whilehaving little or no effect on activation of Akt (S473) and itsdownstream effectors PRAS40 and FOXO1/3 A. Conversely, Aktinhibitors did not modify phosphorylation levels of EGFR orERK, but significantly decreased Akt pathway activation.Both GSK2141795 and AZD5363 completely blocked PRAS40phosphorylation, while only AZD5363 suppressed FOXO1/3Aactivation in the resistant cell lines. Both inhibitorscaused increased Akt phosphorylation in all cell lines, as shownpreviously35, 36, but the reduced phosphorylation levels ofPRAS40 and FOXO1/3A clearly indicated inhibition of Akt

EGFR

pEGFR

AKT

pAKT - S473

PRAS40

pPRAS40

FOXO3

pFOXO1/3a

PTEN

ERK1/2

pERK1/2

Beta actin

EGFR

pEGFR

AKT

pAKT - S473

PRAS40

pPRAS40

FOXO3

pFOXO1/3a

PTEN

ERK1/2

pERK1/2

Beta actin

EGFR

pEGFR

AKT

pAKT - S473

PRAS40

pPRAS40

FOXO3

pFOXO1/3a

PTEN

ERK1/2

pERK1/2

Beta actin

EGFR TKIGSK2141795

PC

9P

C9

PC

9P

C9

PC

9-E

RP

C9-

ER

PC

9-E

RP

C9-

ER

– + – + – + – +– – + + – – + +

EGFR TKIAZD5363

PC

9P

C9

PC

9P

C9

PC

9-E

RP

C9-

ER

PC

9-E

RP

C9-

ER

– + – + – + – +– – + + – – + +

EGFR TKIGSK2141795

PC

9-G

R1

– + – +– – + +

PC

9-G

R1

PC

9-G

R1

PC

9-G

R1

MW(kDa)

35

35

70

70

55

35

35

55

55

250

250

35

a b c

Fig. 4 Effect of EGFR TKIs and Akt inhibitors on key signal transduction proteins in parental and resistant PC9 cells. a Western blot analysis of PC9 andPC9-ER treated for 4 h with erlotinib (30 µM), GSK2141795 (2.5 µM) or the combination; b Western blot analysis of PC9 and PC9-ER treated for 4 h witherlotinib (30 µM), AZD5363 (35 µM), or the combination; c Western blot analysis of PC9-GR1 treated for 4 h with gefitinib (5 µM), GSK2141795 (2.5 µM),or the combination

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6

6 NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications

kinase activity. Combining an EGFR TKI with eitherGSK2141795 or AZD5363 resulted in simultaneous inhibitionof EGFR, PRAS40 and FOXO1/3A phosphorylation (Fig. 4a–c).We also considered other targets of the Akt pathway and included

a PI3K-mTOR inhibitor, GSK2141458, to be combinedwith EGFR inhibitors (Supplementary Fig. 5c). However, lesssynergism was observed with the PI3K-mTOR inhibitor com-pared to the Akt inhibitor GSK2141795 in combination with an

MW (kDa)

MW (kDa)–

+++++++ –

++–––+–+–––+ ––+–––+–+––+ +

11–1

8 G

R4

11–1

8 G

R6

11–1

8 G

R5

11–1

8 G

R4

11–1

8 G

R3

11–1

8 G

R2

11–1

8 G

R1

11–1

8 11

–18

11–1

8 G

R4

11–1

8 G

R4

11–1

8 G

R4

11–1

8 G

R4

11–1

8 G

R4

11–1

811

–18

11–1

811

–18

11–1

811

–18

––+–++––+–

250

GefitinibGSK2141795AZD5363EGFR

pEGFR

AKT

pAKT

PRAS40

pPRAS40

FOXO3a

pFOXO1/3a

ERK1/2

pERK1/2

Actin

EGFREGFR TKI

METFGFR1AXL MER

**

**

*86420

15

10

5

0

Rel

ativ

em

RN

A e

xpre

ssio

n

2.52.01.51.00.50.0

pEGFR

AXL

MET

pMET

Src

pSrc

AKT

pAKT S473

PTEN

PRAS40

pPRAS40

FOXO3a

pFOXO1/3a

ERK1/2

pERK1/2

Actin

250

250

250

130

130

13055

55

55

55

55

55

55

35

35

35

35

70

70

70

70

35

35

35

35

35

35

Resistant to EGFRi

Resistant to EGFRi

2.52.01.51.00.50.0

11–18 GR1 11–18 GR211–18

Flu

ores

cenc

e

30,000

20,000

10,000

0

Flu

ores

cenc

e

30,000

20,000

10,000

0

No dr

ug

EGFRi

AKTi (G)

EGFRi + A

KTi (G)

EGFRi + A

KTi (A)

AKTi (A)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (A)

EGFRi + A

KTi (G)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (A)

EGFRi + A

KTi (G)

No dr

ug

EGFRi

AKTi (G)

AKTi (A)

EGFRi + A

KTi (A)

EGFRi + A

KTi (G)

11–18 GR4

* **

*

**

**

**

** *

* * **

*

* **

*

* **

*

**

11–18 GR5 11–18 GR6

4 Days7 Days

11–18 GR3

11–1

8 G

R1

11–1

8 G

R2

11–1

8 G

R3

11–1

8 G

R4

11–1

8 G

R5

11–1

8 G

R6

11–1

8

11–1

8 G

R1

11–1

8 G

R2

11–1

8 G

R3

11–1

8 G

R4

11–1

8 G

R5

11–1

8 G

R6

11–1

8

11–1

8 G

R1

11–1

8 G

R2

11–1

8 G

R3

11–1

8 G

R4

11–1

8 G

R5

11–1

8 G

R6

11–1

8

11–1

8 G

R1

11–1

8 G

R2

11–1

8 G

R3

11–1

8 G

R4

11–1

8 G

R5

11–1

8 G

R6

11–1

8

a

b d

c

Fig. 5 Expression of selected proteins in parental and resistant 11–18 cells and synergistic effects of EGFR TKIs and Akt inhibitors. a Relative mRNAexpression of selected genes. Experiments were at least performed in triplicates and results are shown as mean± SD. Asterisks indicate significant differencefrom parental PC9 in a Student’s t-test (p< 0.05). b Western blot analysis of selected proteins after 4 h stimulation with 5 µM gefitinib. c Concentrations ofdrugs were chosen to be sub-inhibitory to explore the synergistic potential: gefitinib (EGFRi) 10 µM (except for 18-11 where 100 nM was used); GSK2141795(G) 2.5 µM; AZD5363 (A) 35 µM. All experiments were conducted in seven replicates and results are shown as mean± SD. Asterisks indicate significantdifference in ANOVA one-way test (p< 0.05) for the drug combination-treated cells compared to cells treated with either drug alone at the same time point.d Western blot analysis of 11–18 and 11–18 GR4 treated for 4 h with gefitinib (10 µM), GSK2141795 (2.5 µM) or the combination

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6 ARTICLE

NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications 7

EGFR inhibitor, indicating that Akt inhibition may be the mosteffective strategy for blocking the pathway and survival in thesecells (Supplementary Table 5).

Activation of Akt in 11–18-derived EGFR-TKI-resistant celllines. To validate our findings in an additional and independentsystem, we generated and characterized six other EGFR-TKI-resistant cell lines derived from the 11–18 NSCLC cell line,which harbors the other major EGFR-TKI-sensitizing EGFRmutation observed in NSCLC patients, EGFR L858R. The IC50sfor gefitinib were 40–200-fold higher in the EGFR-TKI-resistant,derivative sub-lines (11–18 GR1-GR6), all of which retained theL858R mutation (Supplementary Table 6). The T790M mutationdid not appear in any of the resistant cell lines. The resistant celllines were morphologically different from the 11–18 parentalcells, with four exhibiting a fusiform appearance (SupplementaryFig. 6). No MET amplification was detected in any of thecell lines. Protein and gene expression analyses revealed AXLupregulation in one, upregulation of FGFR1, MET, and Mer insome, and upregulation of EGFR in all of the 11–18 resistant cells(Fig. 5a and Supplementary Fig. 5). HER2 and HER3 levels

were not elevated in the resistant cell lines compared to 11–18(Supplementary Fig. 5). Low levels of pEGFR were observed inthe presence of gefitinib in all cases. Accordingly, significant levelsof pERK and pSrc were also observed in most of them (Fig. 5b).Finally, high E-cadherin and beta-catenin expression levels wererevealed in the parental cells and were only slightly reduced in theresistant clones. Vimentin expression levels were low in theparental and the resistant cells and no N-cadherin expression wasdetected in any of the cells (Supplementary Fig. 6a). In conclu-sion, resistance mechanisms in the 11–18 cells were diverse anddiffered from those in the PC9 model. No T790M mutations orindication of EMT were detected, and only one cell line showedAXL overexpression, while EGFR was upregulated in all cases.Consequently, while EGFR TKIs completely inactivated theEGFR/ERK pathway in T790M-negative resistant cells derivedfrom PC9, they failed to block activation of ERK and did notcompletely inhibit EGFR phosphorylation in five of the sixresistant cell lines derived from 11–18 cells. Remarkably,increased activation of the Akt pathway was found to be the onlycommon trait between the total of eleven resistant cell lines inthese two cell line models.

PC9-GR4PC9-GR4

PC9-GR4 PC9-GR4

2 Days4 Days6 Days

Rel

ativ

e to

unt

reat

ed (

%)

Rel

ativ

e to

unt

reat

ed (

%)

100

50

0

Rel

ativ

e to

unt

reat

ed (

%)

100

50

0

100

50

0

Rel

ativ

e to

unt

reat

ed (

%)

100

50

0

Vehicl

e

AZD9291

0.1

μM

AZD9291

0.2

μM

GSK2141

795

2.5μM

AZD9291

0.1

μM +

GSK21

4179

5

AZD9291

0.2

μM +

GSK21

4179

5

Vehicl

e

AZD9291

0.1

μM

AZD9291

0.2

μM

AZD9291

0.2

μM +

AZD53

63

AZD5363

35μM

AZD9291

0.1

μM +

AZD53

63

Vehicl

e

Vehicl

e

AZD9291

0.1

μM

AZD9291

0.2

μM

GSK2141

795

2.5μM

AZD9291

0.1

μM +

GSK21

4179

5

AZD9291

0.2

μM +

GSK21

4179

5

AZD9291

0.1

μM

AZD9291

0.2

μM

AZD5363

35μM

AZD9291

0.1

μM +

AZD53

63

AZD9291

0.2

μM +

AZD53

63

* ***

** **

a

b

Fig. 6 Synergistic effect of third-generation EGFR-TKI combined with an Akt inhibitor in T790M-mutated EGFR-TKI-resistant PC9 cells. The combinedeffects of the third-generation EGFR inhibitor (Osimertinib, AZD9291) targeting the T790M resistance mutation and Akt inhibitors (GSK2141795 orAZD5363) on T790M-mutated gefitinib-resistant PC9-GR4 cells were assessed by a CellTiterBlue assay after 2, 4 and 6 days and b BrdU incorporationassay performed after 5 days. Concentrations of drugs were chosen to be sub-inhibitory to explore the synergistic potential: AZD9291 0.1 and 0.2 µM;GSK2141795 2.5 µM; and AZD5363 35 µM. All experiments were conducted in 7 replicates (a), or quadruplicates (b), and results are shown as mean± SD.Asterisks indicate significant difference in ANOVA 1-way test (p< 0.05) for the drug combination-treated cells compared to cells treated with either drugalone. For the CellTiterBlue assay significance differences was observed for days 4 and 6

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6

8 NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications

Synergistic growth inhibition of 11–18 resistant cell lines. Thegrowth inhibitory effect of the combination of gefitinib witheither of the two Akt inhibitors, GSK2141795 and AZD5363, ondifferent 11–18 resistant cell lines was evaluated using CellTi-terBlue (Fig. 5c) and BrdU assays (Supplementary Fig. 7). Asynergistic growth inhibitory effect was observed in all of the11–18 EGFR-TKI-resistant cell lines, but not in the parental11–18 cell line, where the two drugs showed less effect (Fig. 5cand Supplementary Table 4). Similar synergistic growthinhibitory effect on the resistant cell lines was also observed atlower more physiologic concentrations (Supplementary Fig. 8).Western blotting demonstrated a complete inhibition of FOXO1/3a phosphorylation only for the combination of EGFR TKI andAkt inhibitor, whereas PRAS40 phosphorylation was also blockedby the Akt inhibitors as single agent (Fig. 5d). Similarly to whatwas observed in the PC9 panel, both Akt inhibitors led to anincrease in pAKT S473, but still completely prevented phos-phorylation of Akt downstream targets, indicating a lack of Aktkinase functional output.

Third-generation EGFR TKI and Akt inhibitor co-treatment.To evaluate whether synergistic effect on growth between anEGFR TKI and an Akt inhibitor could also be observedwhen using a third-generation EGFR-TKI targeting the T790Mresistance mutation, we assessed the effects of osimertinib witheither of the Akt inhibitors, GSK2141795 and AZD5363. A strongsynergistic growth inhibition of the T790M-mutated EGFR-TKI-resistant cell line PC9-GR4 was observed for both combinationsas determined by CellTiterBlue and BrdU incorporation assays(Fig. 6). Further, we examined whether the combination ofosimertinib and an Akt inhibitor could delay the emergence ofresistance in the T790M-positive PC9-GR4 cells compared to thesingle agents in a colony outgrowth assay. A delay of morethan 8 weeks for the combination treatment was observed(Supplementary Fig. 9). Although slightly lower pEGFR levelswere present in the resistant compared to parental cells in theabsence of gefitinib, EGFR inhibition was critical for growthsuppression in these resistant cell lines, as shown by the

synergistic effect of EGFR TKI and Akt inhibitor co-treatmentstudies (Figs. 3, 5–7 and Supplementary Fig. 9).

EGFR TKI and Akt inhibitor co-treatment of xenograft mice.Next, we evaluated the antitumor activity of combined EGFR TKIand Akt inhibitor treatment in a xenograft model. CB17 SCIDmice were subcutaneously inoculated in both flanks with 1.5x106

PC9-ER cells that harbor increased pAkt and when palpable, micewere administered erlotinib (25 µg/g bodyweight), GSK2141795(10 µg/g bodyweight), the drug combination, or vehicle by oralgavage 5 days a week. Weekly measurements using calipersshowed extensive growth of the vehicle-treated tumors, arelatively modest antitumor effect of erlotinib or GSK2141795alone, and a significant inhibitory effect of the drug combination(Fig. 6a). After 3 weeks of treatment, mice were killed and tumorswere resected. A significant reduction in tumor volume wasobserved between the vehicle group and the group receivingcombination therapy (p= 0.0017), while no significant differenceswere found between the control group and those treated witherlotinib or GSK2141795 as single agents. Significant differenceswere also observed when comparing the tumors receiving thecombination therapy with those treated with erlotinib (p= 0.03)or GSK2141795 (p= 0.0295) alone (Fig. 7b and SupplementaryFig. 10a, b).

pAkt as a biomarker in EGFR-TKI-treated EGFR-mutantpatients. To validate the clinical relevance of our findings, wefirst evaluated pAKT S473 expression by IHC in baselinesamples of 75 EGFR-mutant NSCLC patients treated withfirst-line EGFR-TKIs (Supplementary Table 7). Importantly, thestability of pAkt staining in IHC analysis was found to be robustindependent of fixation time (Supplementary Fig. 11). pAktexpression was determined by a weighted histoscore methodalso known as the H-score37, which takes into account thepercentage of positive signal and staining intensity. To definethe optimal cut-point for pAkt H-score as a continuous variablewe applied the method of Contal and O’Quigley38 that useslog-rank test statistic to estimate the cutpoint39. Using thisapproach 62 was identified as the optimal cut-point divided

Tot

al tu

mor

vol

ume

(mm

3 )

Tot

al tu

mor

volu

me

(mm

3 )

Tot

al tu

mor

volu

me

(mm

3 )

Tot

al tu

mor

volu

me

(mm

3 )

Tot

al tu

mor

volu

me

(mm

3 )500

400

300

200

100

0

Vehicl

e

GSK2141

795

Erlotin

ib

Erlotin

ib +

GSK2141

795

400

600

200

0

Time (days)Time (days)0 10 20 300 10 20 30

Time (days)Time (days)0 10 20 300 10 20 30

Vehicle GSK2141795

Erlotinib + GSK2141795

Total tumor volume

p = 0.0017

p = 0.0295

0.03

Erlotinib

400

600

200

0

400

600

200

0

400

600

200

0

a b

Fig. 7 Growth co-inhibition of EGFR-TKI-resistant PC9-ER tumors in mice treated with EGFR TKI and Akt inhibitor. a Time course assessment of total tumorvolume. Each mouse harbored two tumors, and total tumor volume refers to total volume of two tumors per mouse. PC9-ER cells were inoculated on day 0and treatment was started on day 15. Erlotinib (25 µg/g bodyweight), GSK2141795 (10 µg/g bodyweight) or vehicle was administered by oral gavage oncedaily for 5 days a week, and mice were killed on day 36. b Total tumor volumes measured after excision on day 36. Tumor volumes were calculated by theformula (length×weight2)/2. Results are shown as mean± SD. Asterisks indicate significant difference in ANOVA one-way test (p< 0.05)

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6 ARTICLE

NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications 9

patients into two groups: pAkt negative (histoscore< 62)and pAkt positive (histoscore≥ 62). Eight out of 75 (11%) pre-treatment samples were pAkt positive (Fig. 8a). PFS to EGFR-TKItreatment was 14.5 months (95% confidence interval (CI),12.3–17.9) for pAkt negative patients vs. 6.1 months (95% CI,2.3–15.6) for pAkt-positive patients (p= 0.0037), (hazardratio (HR) 0.341; 95% CI, 0.159–0.731; p= 0.0057; Fig. 8b).Overall survival (OS) was 34.5 months (95% CI, 22.3–39.2) forpAkt negative patients vs. 15.2 months (95% CI, 3.0–24.9) forpAkt-positive patients (p= 0.0015), (HR 0.276; 95% CI, 0.118 to0.646; p= 0.003; Fig. 8c). A univariate Cox proportional hazardsmodel for PFS and OS was fitted using the following clinicallyinteresting variables: type of EGFR mutation, gender, smokinghistory, age, brain metastases, and pAkt. Only the status of pAktcontributed significantly to PFS and OS. For a subpopulation ofthe baseline samples where additional tissue was available weevaluated whether pAkt expression correlated with individualresistance mechanisms by examining the AXL, MET, Her2, andFGFR1 expression and PIK3CA mutation status; however, nosuch correlations were observed (Supplementary Table 8and Supplementary Fig. 12) and high pAKT levels, althoughinfrequent, were observed across the tumors with each of theseknown resistance mechanisms. These data reveal a potentiallyimportant role for increased pAkt levels as a novel biomarker topredict decreased initial EGFR-TKI response in patients acrossthe spectrum of previously established individual mechanisms ofresistance.

We next used IHC to evaluate pAkt in rebiopsy samples from15 EGFR-mutant NSCLC patients after progression to EGFR TKI.Nine of those 15 (60%) samples were pAkt positive (Fig. 8aand Supplementary Table 9), compared with 11% pAkt

positive in pre-treatment samples. Representative images ofpre- and post-treatment samples stained for pAkt are depictedin Fig. 8b. We furthermore evaluated common resistancemechanisms to EGFR TKI in the 15 post-treatment samples,and found that they represented a wide array of well-knownresistance mechanisms (Supplementary Table 10), includingT790M mutation, PIK3CA mutation, AXL upregulation,MET amplification, or upregulation and alterations in markersof EMT, similarly to what we have observed in our wide variety ofour cell line models with acquired EGFR-TKI resistance.Increased pAkt was observed across clinical specimens that havevarious established resistance-associated alterations (e.g., T790M,high AXL, high MET, and EMT), consistent with ourpreclinical findings. Importantly, increased pAkt was identifiedindependent of T790M status, indicating that these alterationsare neither mutually exclusive nor exclusively coupled in theseEGFR-TKI-resistant tumors. Our findings in these clinicalsamples provide evidence for a previously unappreciatedconvergent role of Akt activation in acquired EGFR-TKIresistance that is associated with otherwise diverse andunpredictable upstream molecular events.

DiscussionNSCLC patients with activating EGFR mutations benefit fromtreatment with EGFR TKIs, but ultimately acquire resistance,which limits PFS to 9–13 months and prevents long-term patientsurvival2. Multiple mechanisms of acquired resistance to EGFRTKI have been described by us and other investigators10, 12–16.This diversity, coupled with the unpredictability, of EGFR-TKIresistance mechanisms presents a major challenge for developing

pAK

T

Pre EGFR TKI Post EGFR TKI

Pro

gres

sion

-fre

e-s

urvi

val

Ove

rall

surv

ival

Time (months)

0 9 15 21 27 33 39 45 51 57 63

Time (months)

0 9 15 21 27 33 39 45 51 57 63

Patients at riskpAKT negative

pAKT positive67 63 58 45 42 32 25 23 21 19 16 8 8 4 4 4 3 3 2 2 2 0 0

8 7 4 2 2 2 1 0 00 0 0 0 0 0 0 0 0 0 0 0 0 0

Patients at riskpAKT negative

pAKT positive67 65 61 58 55 48 42 39 34 32 25 21 17 13 10 9 7 5 4 4 3 1 0

8 8 6 6 4 4 4 0 02 1 0 0 0 0 0 0 0 0 0 0 0 0

pAKT negative

pAKT positivepAKT

post-treatmentpAKT

pre-treatment

11%

89%

60%40%

b

1.00

0.75

0.50

0.25

0.00

1.00

0.75

0.50

0.25

0.00

pAkt positive

pAkt negative

pAkt positive

pAkt negative

a

c

Fig. 8 Assessment of pAKT in clinical samples. a Pie chart (left) shows percentage of pAKT-positive or -negative EGFR-mutant NSCLC patients at baselinetreated with first-line EGFR-TKIs. Pie chart (right) shows percentage of pAKT-positive or -negative EGFR-mutant NSCLC patients progressing to first-lineEGFR-TKIs. The histoscore value of 62 was used to classify samples as positive (above 62) or negative (below 62) for pAKT. b Representative negativeand positive immunoshistochemical staining of pAKT. Scale bar, 100 µm. c Left: progression-free survival according to baseline pAKT expression for75 patients with EGFR-mutant NSCLC treated with first-line EGFR-TKI. Median progression-free survival was 14.5 months (95% CI 12.3–17.9) forthe 67 pAKT-negative patients (green line) and 6.1 months (95% CI 2.3–15.6) for the 8 pAKT-positive patients (pink line); p= 0.0037; HR= 0.341;95% CI 0.159–0.731; p= 0.0057. Right: OS according to baseline pAKT expression for 75 patients with EGFR-mutant NSCLC treated with first-lineEGFR-TKI. Median OS was 34.5 months (95% CI 22.3–39.2) for the 67 pAKT-negative patients (green line) and 15.2 months (95% CI 3.0–24.9) forthe 8 pAKT-positive patients (pink line); p= 0.0015; HR= 0.276; 95% CI 0.118–0.646; p= 0.003

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6

10 NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications

effective treatment regimens to overcome EGFR-TKI resistance.Here, we show for the first time (to our knowledge) that suchdiversity in acquired resistance mechanisms is associated withconvergent activation of the Akt pathway in EGFR-mutantNSCLC. Importantly, we have shown that the combination of anEGFR TKI and an Akt inhibitor elicits synergistic growthinhibition in vitro and significant growth inhibition in vivo inotherwise EGFR-TKI-resistant NSCLCs. We propose that byspecifically inhibiting Akt together with EGFR, these otherwisediverse resistance mechanisms can be effectively controlled,allowing for a more rational and feasible treatment strategyto address both the molecular heterogeneity of EGFR-TKIresistance40 and the simultaneous presence of more than onemechanism of resistance (Supplementary Table 8) that wenow know characterizes EGFR-TKI progression in many EGFR-mutant NSCLC patients. The therapeutic strategy identified inour study here provides an alternative approach that maymore efficiently combat the profound molecular heterogeneitythat is an obstacle to the success of current treatments that seek toblock an individual upstream resistance mechanism (for instance,an RTK) to overcome acquired EGFR-TKI resistance.

The importance of the Akt pathway in EGFR-TKI resistancewas validated in clinical samples from EGFR-mutant NSCLCpatients, where phosphorylation of Akt was observed in 60% oftumor samples from patients after progression on EGFR TKIs,but only in 11% of baseline samples. Positivity of pAkt inpost-progression samples was found independent of T790Moccurrence. Moreover, the presence of pAkt positivity in 8 out of75 baseline samples correlated with significantly worse PFSand OS to first-line EGFR-TKI treatment (6.1 vs. 14.5 months;p= 0.0037 and 15.2 vs. 34.5 months; p= 0.0015, respectively).Our clinical data provide new evidence for the potentialclinical utility of assessing pAkt levels as a molecular predictor ofEGFR-TKI response and resistance in EGFR-mutant NSCLCpatients.

Activation of the PI3K-Akt pathway has been reported inEGFR-mutant cell lines and baseline tumors sensitive to TKIsand some reports have associated it with intrinsic sensitivity tofirst-line treatment with EGFR TKIs28–32, 41, 42. However, ourreport is the first to highlight the importance of Akt activation incell lines or tumors with acquired resistance to EGFR TKIs and,most importantly, across tumors with a multitude of previouslyestablished resistance-promoting mechanisms.

Three EGFR inhibitors, erlotinib and gefitinib, and osimertinib,and two Akt inhibitors, GSK2141795 and AZD5363, wereemployed in our study to ensure the validity of the observations.We did not attempt knock-down of Akt to verify the effect of theAkt inhibitors, as no single probe can target all three isoforms ofAkt simultaneously due to sequence heterogeneity. GSK2141795and AZD5363 are orally available inhibitors of Akt1-3 withnanomolar or sub-nanomolar potency35, 43. Combined EGFRTKI and Akt inhibitor treatment of the EGFR-TKI-resistantNSCLC cells generally induced a complete inhibition ofphosphorylation of the two downstream targets PRAS40 andFOXO1/3A. FOXO1/3A is a key pro-apoptotic protein, which,upon phosphorylation by Akt, interacts with 14-3-3 in thenucleus and is transported to the cytoplasm. PhosphorylatedFOXO1/3A cannot re-enter the nucleus to trigger transcription ofthe pro-apoptotic program23. PRAS40 functions as a negativeregulator of mTORC1, which is a complex involved in proteintranslation and ribosome biogenesis. Upon phosphorylationby Akt, PRAS40 becomes inhibited and can no longer blockactivation of the translational machinery and protein synthesis25.Interestingly, both Akt inhibitors elicited a feedbackhyperphosphorylation of Akt at Ser473, as observed by otherinvestigators35, 36. This differs from the effect observed when

employing PI3K or mTOR inhibitors, which result in a decreasein pAkt Ser47344–46.

Prior studies have evaluated the efficacy of Akt inhibitors in arange of cancers, where promising results had been obtained inpreclinical models using drug combinations that include Aktinhibitors, such as AZD5363 combined with fulvestrant inendocrine-resistant breast cancer47, and AZD5363 combined withAZD8931, an EGFR/ErbB2/ErbB3 inhibitor, in ErbB2-amplifiedbreast cancer48. Enhanced tumor growth delay was also observedfor GSK2141795 in combination with a MEK inhibitor in modelsof pancreatic cancer35. In the clinic, erlotinib has been combinedwith the non-ATP competitive pan Akt inhibitor MK-2206 in aphase II clinical trial enrolling advanced NSCLC patients witheither mutant or wildtype EGFR49. In contrast to our findingsreported here, the rationale for this trial was not to blockincreased Akt activation present in the tumors during EGFR-TKItreatment, as pAkt expression was not examined in these patients,but instead to mitigate hepatocyte growth factor-mediatedresistance. In the EGFR-mutant arm of the trial, eligiblepatients had earlier benefited from, but since progressed on,erlotinib as a single agent. Of 45 EGFR-mutant patients, four hada partial response and 14 had stable disease. Importantly, the levelof pAkt was not evaluated in the patients prior to treatment, andwe have shown in our study that cell lines with high levels ofpAKT are most responsive to combined EGFR and Aktinhibition. However, we have not evaluated resistant celllines with a low level of pAkt. Based on our new findings, wehypothesize that there would be a correlation between the level ofpAKT and the efficacy of the erlotinib and MK-2206 in thepatients from the study by Lara and colleagues49, but this infor-mation has not been reported nor have pAKT levels beenexamined retrospectively through molecular studies in thesetumor specimens to our knowledge. Thus, our findings nowprovide the rationale to do so and pave the way for novelbiomarker-driven clinical trials testing an EGFR-TKI plus an Aktinhibitor in appropriately selected EGFR-mutant NSCLC patientswith high pAkt tumor expression.

In conclusion, we have shown that Akt pathway activation iscommonly associated with acquired resistance to EGFR-TKItreatment in NSCLCs harboring a diverse array of other,previously identified resistance mechanisms. Phosphorylation ofAkt was detected in a minority of EGFR-mutant NSCLCpatients prior to EGFR-TKI therapy and predicted worse initialEGFR-TKI response. Finally, we have demonstrated thatcombined treatment with Akt and EGFR inhibitors elicitedsynergistic growth inhibition in preclinical models of EGFR-TKIresistance. Our findings reveal the unanticipated convergentactivation of, and dependence on, Akt in EGFR-TKI resistancethat emerges through a multitude of diverse upstream events.While other studies have tested inhibitors of PI3K and/or mTORin EGFR-mutant NSCLC, our data provide a novel rationale forspecifically testing Akt inhibitors in combination with EGFRTKIs in EGFR-mutant NSCLC patients that show Akt activation.We propose that Akt inhibition, specifically, could moreuniformly enhance response and survival in patients with highpAkt levels who are at high risk for Akt-mediated resistance, asthis distinct approach has the unique potential to combat theotherwise profound heterogeneity of molecular resistance eventsthat are present in EGFR-mutant NSCLC patients with acquiredEGFR-TKI resistance to improve their outcomes.

MethodsCell culture and antitumor agents. Parental PC9 cells were kindly provided byF. Hoffman-La Roche Ltd (Basel, Switzerland) with the authorization of Dr.Mayumi Ono (Kyushu University, Fukuoka, Japan). Parental 11–18 cells werekindly provided by Dr. Mayumi Ono. All cell culture materials were obtained from

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6 ARTICLE

NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications 11

Biological Industries (Kibbutz Beit Haemek, Israel) or Gibco Life Technologies(Carlsbad, CA, USA). Cells were cultured in a humidified atmosphere with 5% CO2

at 37 °C in RPMI1640 + 10% fetal bovine serum (FBS), 50 µg/mL penicillin-streptomycin and 2 mM L-Glutamine. EGFR inhibitors (erlotinib, gefitinib,afatinib, dacomitinib, and osimertinib (AZD9291)) and the Akt inhibitorAZD5363 were purchased from Selleck Chemicals (Houston, TX, USA).The Akt inhibitor GSK2141795 (Uprosertib) was purchased from MedchemexpressLLC (Monmouth Junction, NJ, USA). All drugs were dissolved in DMSO, aliquotedand kept at −20 °C.

Mutation analyses. DNA was isolated from cell lines and hematoxilin/eosinor IHC slides by standard procedures and genotyped by quantitative PCR withspecific probes (Taqman) or standard PCR followed by Sanger sequencing.

Gene expression analyses. RNA was isolated from the cell lines in accordancewith a proprietary procedure (European patent number EP1945764-B1) aspreviously described50. The primer and probe sets were designed using PrimerExpress 3.0 Software (Applied Biosystems, Foster City, CA, USA) according totheir Ref Seq (http://www.ncbi.nlm.nih.gov/LocusLink). Quantification of geneexpression was performed using the ABI Prism 7900HT Sequence DetectionSystem (Applied Biosystems). Expression levels were calculated according to thecomparative ΔΔCt method. Commercial RNA controls were used as calibrators(Liver and Lung; Stratagene, La Jolla, CA, USA). For each cell line, threeindependent experiments were performed.

Western blotting. Cells were washed in ice-cold tris buffered saline (TBS), spundown and lysed in radioimmunoprecipitation assay buffer (10 mM Tris HCl, pH 8,5 mM Na2EDTA, pH 8, 1% NP-40, 0.5% sodium dioxycholate, 0.1% SDS), bothcontaining protease and phosphatase inhibitors (Complete Mini PhosphoSTOP,Roche, Basel, Switzerland). Protein concentrations were determined by Pierce BCAProtein Assay (Thermo Fisher Scientific, Rockford, IL, USA) according to themanufacturer’s protocol. 5–40 μg protein was resolved on 4–12% RunBlueSDS-PAGE gels (Expedeon, San Diego, CA, USA), transferred onto PVDFmembrane (GE Healthcare Life Sciences, Buckinghamshire, UK), blocked andthen incubated with primary monoclonal antibodies ON at 4 °C. Following, themembranes were incubated with goat anti-rabbit, goat anti-mouse (#P0448,#P0447, Dako, Glostrup, Denmark) or donkey anti-goat (#SC-2020, Santa CruzBiotechnology, Heidelberg, Germany) HRP-conjugated secondary antibodies in1:5000 dilution for 1 h at room temperature (RT). The immune reactive bandswere visualized by Amersham ECL Prime Western Blotting Detecting Reagent(GE Healthcare Life Sciences) and exposed to CL-Xposure film (Thermo FisherScientific).

Immunohistochemistry. IHC was performed on 5 μm sections using an auto-mated immunostainer (Ventana BenchMark ULTRA, Ventana Medical Systems,Oro Valley, AZ, USA) and protein expression was quantified using the histoscoremethod as previously described51. The following antibodies were used: AXL(Cell Signaling Technology, Leiden, The Netherlands, #8661, dilution 1:100),pAkt (Cell Signaling Technology, #4060, dilution 1:50), phospho-PRAS40(Cell Signaling Technology, #2997, dilution 1:100), E-cadherin (Roche#5973872001, RTU), β-catenin (Roche #5269016001, RTU), vimentin (Roche#5278139001, RTU), and N-cadherin (Abcam, Cambridge, UK, #ab18203, dilution1:100).

Cell growth and apoptosis assays. Cells were seeded at 4000 cells per well in 96-well plates, allowed to attach for 24 h and treated for 72 h with drug. After treat-ment, cells were incubated with medium containing MTT (0.75 mg/mL in med-ium) for 1 h at 37 °C. Culture medium with MTT was then removed and formazancrystals dissolved in 100 µL DMSO (Sigma-Aldrich, St. Louis, MO, USA). Cellnumbers were estimated by measuring the absorbance at 495 nm using a micro-plate reader (BioWhittaker, Walkersville, MD, USA).

For drug combination experiments, 10,000 cells per well were seeded in 24-wellplates and left to attach for 6 h before drugs or vehicle were added; then grown at37 °C. Cell viability was quantified by crystal violet staining or by CellTiterBlue(Promega, WI, USA) and cell proliferation was evaluated by BrdU incorporationusing the BrdU Cell Proliferation Assay Kit (Cell Signaling Technology, Beverly,MA, USA), according to the manufacturer´s instructions. Crystal violet assay wasperformed by adding staining solution for 5 min at RT, washing cells twice in H2O,redissolving inNa-citrate buffer (29.41 g Na-citrate in 50% EtOH) and measuring the absorbanceat 570 nm. Apoptosis was assessed using the Cell Death Detection ELISAPlus kit(Roche, Basel, Switzerland) according to the manufacturer’s instructions. Colonyoutgrowth assay was performed as described in Tricker et al. (2015)52.

SILAC labeling. Cells were propagated in RPMI1640 (BioNordika, Herlev,Denmark) supplemented with 580 mg/L-glutamine and 200 mg/L proline and 10%dialyzed FBS (Gibco, Life Technologies, Carlsbad, CA, USA). PC9-ER cells werepropagated in “heavy” media containing 75 mg/L 13C6-Lys and 28 mg/L 13C6-Arg

(Cambridge Isotope Laboratories, Tewksbury, MA, USA), and PC9 cells werepropagated in “light” medium containing 75 mg/L 12C6-Lys and 28 mg/L 12C6-Arg(Sigma-Aldrich). Cells were propagated for six passages in SILAC media andharvested to ascertain complete isotope incorporation.

Cell harvest for mass spectrometry. All cells were harvested at 80% confluence.Cells were washed twice with ice-cold TBS and lyzed with ice-cold 0.1 M Na2CO3

pH 11 lysis buffer containing 1 mM activated sodium pervanadate andprotease (Complete Mini EDTA-free, Roche, Basel, Switzerland) and phosphatase(Complete Mini PhosSTOP, Roche, Basel, Switzerland) inhibitors. Cell scraperswere used not to damage extracellular membrane proteins. The lysate was adjustedto 1 mM MgCl2 and 3 µL benzonase (Sigma-Aldrich) was added and sampleswere left on ice for 15 min to degrade RNA and DNA. Lysates were nowhomogenized using a Branson sonifier 250, 2 × 30 s, output 10, output control 2.5.The lysates were then centrifuged at 100,000×g for 45 min at 4 °C in a SorvallRC M150 GX ultracentrifuge to separate soluble proteins (supernatant)from membraneous proteins (pellet). The pellets were washed with 0.5 Mtriethylammonium bicarbonate (TEAB) followed by 0.05 M TEAB to removesoluble protein contamination.

Protein purification and digestion. The supernatant proteins were precipitatedby adding four volumes of ice cold acetone and left at −20 °C followed bycentrifugation at 7000×g for 15 min. The pellets containing soluble proteinswere then dissolved in 8M urea and incubated to fully dissolve protein clusters.Protein concentrations were determined by the Bio-Rad Protein Assay beforeproteins were reduced by 20 mM dithiothreitol for 45 min and subsequentlyalkylated by 40 mM iodoacetamide for 45 min in the dark. Samples were thendigested with Lys-C (Sigma-Aldrich) 1 µg enzyme/50 µg protein for 5 h at RT.Samples were diluted eight times with 0.1 M ammonium bicarbonate and digestedwith trypsin (1 µg enzyme/100 µg peptide) at 37 °C. Samples were acidified to 0.1%trifluoroacetic acid (TFA) and centrifuged at 15,000×g for 10 min to pellet insolublematerials such as lipids, and the supernatant was kept for further analysis. Themembrane proteins were dissolved in 8M urea, reduced and alkylated as statedabove before 0.5 µL Sialidase A (Europa Bioproducts, Cambridge, UK) and 1 µLPNGase F (Sigma-Aldrich) was added at 37 °C to remove extracellular glycanstructures. Then, samples were treated with Lys-C and trypsin as described above.Digested peptides were desalted on in-house packed stage tip columns composed oftwo C18 membrane disks (Empore 3M, Bellefonte, USA) and porous R2/R3reverse-phase resins (Applied Biosystems). In brief, samples were acidified to pH~2 before peptides were applied to 0.1% TFA pre-equilibrated columns, washedwith 0.1% TFA and eluted using 70% ACN, 0.1% TFA.

HILIC fractionation. Samples were adjusted to ~40 μL of 90% ACN, 0.1% TFA byfirst dissolving it in 0.4 μL of 10% TFA, then adding 3.6 μL of H2O, and finallyadding 36 μL of ACN. The samples were injected onto an in-house packed TSKgelAmide-80 HILIC 320 μm× 170mm capillary HPLC column using an Agilent1200 capillary HPLC system. The peptides were eluted using a gradient from 90%ACN, 0.1% TFA to 60% ACN, 0.1% TFA for over 46 min at a flow rate of 6 μL/min.The fractions were automatically collected in a microwell plate at 1 minintervals after UV detection at 210 nm, and the fractions were pooled according tothe UV detection to a total of 10 fractions. The fractions were dried by vacuumcentrifugation. Prior to LC–MS/MS the samples were redissolved in 0.5 μL of 100%formic acid (FA) and diluted with H2O to 5.5 μL. A total of 5 μL of each fractionwere analyzed by reverse-phase nanoLC–MS/MS.

NanoLC–MS/MS and data analysis. The peptides were loaded onto anEasy-nanoLC (Thermo Fisher Scientific) coupled to an Orbitrap FusionTMTribridTM mass spectrometer (Thermo Fisher Scientific). Peptides were loadedonto a pre-column (2 cm Reprosil—Pur C18 AQ 5 µm RP material (Dr. Maisch,Ammerbuch-Entrigen, Germany)) using the EASY-LC system and eluteddirectly onto a 20 cm-long fused silica capillary column (75 µm ID) packed withReprosil—Pur C18 AQ 3 µm RP material. The peptides were separated using agradient from 0–34% B (A buffer: 0.1 % FA; B buffer: 90% ACN/0.1% FA) at aflow rate of 250 nL/min over 30–90 min depending on the UV trace of theHILIC fractions. Peptides (m/z 400–1400) were analyzed in full MS mode using aresolution of 120.000 FWHM at 200 m/z and the peptides were selected andfragmented using high-energy collisional dissociation (HCD) and the fragmentions were recorded in the LTQ with low resolution (rapid scan rate). A maximumof 3 s were allowed between each MS and for MSMS the ion filling time was set to35 ms and an AGC target value of 2E4 ions. The mass spectrometry data wereprocessed using Proteome Discoverer software v1.4 (Thermo Scientific), and theraw data were searched in the Swissprot and Uniprot databases using the Mascotand SEQUEST search algorithms. Quantitation using SILAC was performed inProteome Discoverer v 1.4 using the SILAC Quantitation node.

We calculated the average of the triplicate ratios for each protein and selectedthose that fulfilled two criteria. First, the average ratio should be ≥2-folddifferentially expressed, and second, at least two of the triplicate ratios should be≥2-fold. We then eliminated the cytoplasmatic and membrane proteins that wereonly up or downregulated in the membrane or cytoplasmic fraction, respectively.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6

12 NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications

Ingenuity Pathway Analysis software (Qiagen, Redwood City, CA, USA) was usedfor classifying proteins in functional groups.

Mice xenograft study. All animal experiments were approved by TheExperimental Animal Committee of The Danish Ministry of Justice and wereperformed at the animal core facility at University of Southern Denmark. Micewere housed under pathogen-free conditions with ad libitum food and water.Subconfluent PC9-ER cells (1.5 x 106) were harvested using accutase andresuspended in a 1:1 mixture of extracellular matrix from Engelbreth-Holm-Swarmsarcoma (Sigma-Aldrich) and RPMI1640 media, and injected subcutaneously into16-week-old female CB17 SCID mice (Taconic, Ejby, Denmark). When tumorsreached a palpable diameter of 2–3 mm, mice were randomized into groups of n=9 per group. Animals were euthanized if they showed any adverse signs of diseaseincluding weight loss, paralysis, thymus dysfunction, or general discomfort.Accordingly, eight mice were censored during the course of the study. ErlotinibHCl was formulated at 25 µg/g bodyweight in 15% Captisol (La Jolla, CA, USA).GSK2141795 was formulated at 10 µg/g bodyweight in DMSO. 15% Captisol wasused as vehicle, and the concentration of DMSO did not exceed 10% whenadministered. Drugs were administered 5 days a week for 3 weeks by oral gavage.Maximum volume per mouse was 200 µL. Tumor volume and bodyweight weresurveyed during the extent of the study. Tumor volume was measured with calipersand calculated according to: tumor volume (mm3)= (length×width2)/2.

Patient samples collection. Clinical data were assessed and tumor samplesstudied in accordance with an approved protocol of the institutional review boardof Germans Trias i Pujol Hospital, Badalona, Spain and de-identified for patientconfidentiality.

Statistical methods. PFS and OS were estimated by the Kaplan–Meier method,and the nonparametric log-rank test was used to evaluate differences betweengroups. Cox proportional hazard regression model was applied with pAKT ascovariate, obtaining HR and their 95% CIs. Each analysis was performed withthe use of a two-sided 5% significance level and a 95% CI. Statistical analyseswere performed using SAS version 9.3. Synergy was defined as the two drugstogether having an effect greater than the sum of the two drugs separate effects(CI<1).

Data availability. All relevant data are available from the authors upon request.

Received: 30 October 2016 Accepted: 29 June 2017

References1. Ferlay, J. et al. Cancer incidence and mortality worldwide: sources, methods and

major patterns in GLOBOCAN 2012. Int. J. Cancer 136, E359–E386 (2015).2. Rosell, R. et al. Screening for epidermal growth factor receptor mutations in

lung cancer. N. Engl. J. Med. 361, 958–967 (2009).3. Mitsudomi, T. et al. Gefitinib versus cisplatin plus docetaxel in patients with

non-small-cell lung cancer harbouring mutations of the epidermal growthfactor receptor (WJTOG3405): an open label, randomised phase 3 trial. LancetOncol. 11, 121–128 (2010).

4. Maemondo, M. et al. Gefitinib or chemotherapy for non-small-cell lung cancerwith mutated EGFR. N. Engl. J. Med. 362, 2380–2388 (2010).

5. Zhou, C. et al. Erlotinib versus chemotherapy as first-line treatment for patientswith advanced EGFR mutation-positive non-small-cell lung cancer (OPTIMAL,CTONG-0802): a multicentre, open-label, randomised, phase 3 study. LancetOncol. 12, 735–742 (2011).

6. Rosell, R. et al. Erlotinib versus standard chemotherapy as first-line treatmentfor European patients with advanced EGFR mutation-positive non-small-celllung cancer (EURTAC): a multicentre, open-label, randomised phase 3 trial.Lancet Oncol. 13, 239–246 (2012).

7. Sequist, L. V. et al. Phase III study of afatinib or cisplatin plus pemetrexed inpatients with metastatic lung adenocarcinoma with EGFR mutations. J. Clin.Oncol. 31, 3327–3334 (2013).

8. Wu, Y. L. et al. Afatinib versus cisplatin plus gemcitabine for first-linetreatment of Asian patients with advanced non-small-cell lung cancerharbouring EGFR mutations (LUX-Lung 6): an open-label, randomised phase 3trial. Lancet Oncol. 15, 213–222 (2014).

9. Sordella, R., Bell, D. W., Haber, D. A. & Settleman, J. Gefitinib-sensitizingEGFR mutations in lung cancer activate anti-apoptotic pathways. Science 305,1163–1167 (2004).

10. Pao, W. et al. Acquired resistance of lung adenocarcinomas to gefitinib orerlotinib is associated with a second mutation in the EGFR kinase domain.PLoS Med. 2, e73 (2005).

11. Kobayashi, S. et al. EGFR mutation and resistance of non-small-cell lung cancerto gefitinib. N. Engl. J. Med. 352, 786–792 (2005).

12. Tan, C. S., Gilligan, D. & Pacey, S. Treatment approaches for EGFR-inhibitor-resistant patients with non-small-cell lung cancer. Lancet Oncol. 16, e447–e459(2015).

13. Xie, S. et al. Mer receptor tyrosine kinase is frequently overexpressed in humannon-small cell lung cancer, confirming resistance to erlotinib. Oncotarget 6,9206–9219 (2015).

14. Koch, H., Busto, M. E., Kramer, K., Medard, G. & Kuster, B. Chemicalproteomics uncovers EPHA2 as a mechanism of acquired resistance to smallmolecule EGFR Kinase inhibition. J. Proteome Res. 14, 2617–2625 (2015).

15. Lee, H. J. et al. Drug resistance via feedback activation of Stat3 inoncogene-addicted cancer cells. Cancer Cell 26, 207–221 (2014).

16. Bivona, T. G. et al. FAS and NF-kappaB signalling modulate dependence oflung cancers on mutant EGFR. Nature 471, 523–526 (2011).

17. Turke, A. B. et al. Preexistence and clonal selection of MET amplification inEGFR mutant NSCLC. Cancer Cell 17, 77–88 (2010).

18. Zhang, Z. et al. Activation of the AXL kinase causes resistance toEGFR-targeted therapy in lung cancer. Nat. Genet. 44, 852–860 (2012).

19. Yuan, T. L. & Cantley, L. C. PI3K pathway alterations in cancer: variations on atheme. Oncogene 27, 5497–5510 (2008).

20. Altomare, D. A. & Testa, J. R. Perturbations of the AKT signaling pathwayin human cancer. Oncogene 24, 7455–7464 (2005).

21. Alessi, D. R. et al. Characterization of a 3-phosphoinositide-dependent proteinkinase which phosphorylates and activates protein kinase Balpha. Curr. Biol. 7,261–269 (1997).

22. Sarbassov, D. D., Guertin, D. A., Ali, S. M. & Sabatini, D. M. Phosphorylationand regulation of Akt/PKB by the rictor-mTOR complex. Science 307,1098–1101 (2005).

23. Luo, J., Manning, B. D. & Cantley, L. C. Targeting the PI3K-Akt pathway inhuman cancer: rationale and promise. Cancer Cell 4, 257–262 (2003).

24. Shaw, R. J. & Cantley, L. C. Ras, PI(3)K and mTOR signalling controls tumourcell growth. Nature 441, 424–430 (2006).

25. Wiza, C., Nascimento, E. B. & Ouwens, D. M. Role of PRAS40 in Akt andmTOR signaling in health and disease. Am. J. Physiol. Endocrinol. Metab. 302,E1453–E1460 (2012).

26. Havel, J. J., Li, Z., Cheng, D., Peng, J. & Fu, H. Nuclear PRAS40 couples theAkt/mTORC1 signaling axis to the RPL11-HDM2-p53 nucleolar stressresponse pathway. Oncogene 34, 1487–1498 (2015).

27. Rodon, J., Dienstmann, R., Serra, V. & Tabernero, J. Development of PI3Kinhibitors: lessons learned from early clinical trials. Nat. Rev. Clin. Oncol. 10,143–153 (2013).

28. Amann, J. et al. Aberrant epidermal growth factor receptor signaling andenhanced sensitivity to EGFR inhibitors in lung cancer. Cancer Res. 65,226–235 (2005).

29. Engelman, J. A. et al. ErbB-3 mediates phosphoinositide 3-kinase activity ingefitinib-sensitive non-small cell lung cancer cell lines. Proc. Natl Acad. Sci.USA 102, 3788–3793 (2005).

30. Ono, M. et al. Sensitivity to gefitinib (Iressa, ZD1839) in non-small cell lungcancer cell lines correlates with dependence on the epidermal growth factor(EGF) receptor/extracellular signal-regulated kinase 1/2 and EGF receptor/Aktpathway for proliferation. Mol. Cancer Ther. 3, 465–472 (2004).

31. Fujimoto, N. et al. High expression of ErbB family members and their ligandsin lung adenocarcinomas that are sensitive to inhibition of epidermal growthfactor receptor. Cancer Res. 65, 11478–11485 (2005).

32. Faber, A. C. et al. Differential induction of apoptosis in HER2 and EGFRaddicted cancers following PI3K inhibition. Proc. Natl Acad. Sci. USA 106,19503–19508 (2009).

33. Cheung, H. W. et al. Amplification of CRKL induces transformation andepidermal growth factor receptor inhibitor resistance in human non-small celllung cancers. Cancer Discov. 1, 608–625 (2011).

34. Kohler, A. et al. A hormone-dependent feedback-loop controls androgenreceptor levels by limiting MID1, a novel translation enhancer and promoter ofoncogenic signaling. Mol. Cancer 13, 146 (2014).

35. Dumble, M. et al. Discovery of novel AKT inhibitors with enhanced anti-tumoreffects in combination with the MEK inhibitor. PLoS ONE 9, e100880 (2014).

36. Rhodes, N. et al. Characterization of an Akt kinase inhibitor with potentpharmacodynamic and antitumor activity. Cancer Res. 68, 2366–2374 (2008).

37. Kirkegaard, T. et al. Observer variation in immunohistochemical analysis ofprotein expression, time for a change? Histopathology 48, 787–794 (2006).

38. Contal, C. & O’Quigley, J. An application of changepoint methods in studyingthe effect of age on survival in breast cancer. Comput. Stat. Data Anal 30,253–270 (1999).

39. Mandrekar, J., Mandrekar, S. & Cha, S. Cutpoint determination methods insurvival analysis using SAS1. Presented at: SAS Users Group 28; 30 March–2April 2003; Seattle, WA, USA.

40. Govindan, R. et al. Genomic landscape of non-small cell lung cancer in smokersand never-smokers. Cell 150, 1121–1134 (2012).

NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6 ARTICLE

NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications 13

41. Yamamoto, H. et al. PIK3CA mutations and copy number gains in human lungcancers. Cancer Res. 68, 6913–6921 (2008).

42. Jin, G. et al. PTEN mutations and relationship to EGFR, ERBB2, KRAS, and TP53mutations in non-small cell lung cancers. Lung Cancer 69, 279–283 (2010).

43. Davies, B. R. et al. Preclinical pharmacology of AZD5363, an inhibitor of AKT:pharmacodynamics, antitumor activity, and correlation of monotherapyactivity with genetic background. Mol. Cancer Ther. 11, 873–887 (2012).

44. Rodrik-Outmezguine, V. S. et al. mTOR kinase inhibition causesfeedback-dependent biphasic regulation of AKT signaling. Cancer Discov. 1,248–259 (2011).

45. Hancox, U. et al. Inhibition of PI3Kbeta signaling with AZD8186 inhibitsgrowth of PTEN-deficient breast and prostate tumors alone and in combinationwith docetaxel. Mol. Cancer Ther. 14, 48–58 (2015).

46. Raja, M. et al. Assessment of the in vivo activity of PI3K and MEK inhibitorsin genetically defined models of colorectal cancer. Mol. Cancer Ther. 14,2175–2186 (2015).

47. Ribas, R. et al. AKT antagonist AZD5363 influences estrogen receptor functionin endocrine-resistant breast cancer and synergizes with fulvestrant(ICI182780) In Vivo. Mol. Cancer Ther. 14, 2035–2048 (2015).

48. Crafter, C. et al. Combining AZD8931, a novel EGFR/HER2/HER3 signallinginhibitor, with AZD5363 limits AKT inhibitor induced feedback and enhancesantitumour efficacy in HER2-amplified breast cancer models. Int. J. Oncol. 47,446–454 (2015).

49. Lara, P. N. Jr. et al. Phase II study of the AKT inhibitor MK-2206 plus Erlotinibin patients with advanced non-small cell lung cancer who previously progressedon erlotinib. Clin. Cancer Res. 21, 4321–4326 (2015).

50. Rosell, R. et al. Pretreatment EGFR T790M mutation and BRCA1 mRNAexpression in erlotinib-treated advanced non-small-cell lung cancer patientswith EGFR mutations. Clin. Cancer Res. 17, 1160–1168 (2011).

51. Karachaliou, N. et al. BIM and mTOR expression levels predict outcometo erlotinib in EGFR-mutant non-small-cell lung cancer. Sci. Rep. 5, 17499(2015).

52. Tricker, E. M. et al. Combined EGFR/MEK inhibition prevents theemergence of resistance in EGFR-mutant lung cancer. Cancer Discov. 5,960–971 (2015).

AcknowledgementsWe thank the following investigators for providing us with clinical tumor samples:Andres Felipe Cardona, Clinical and Translational Oncology Group-FICMAC,Colombia; Guillermo Lopez Vivanco, Hospital de Cruces de Barakaldo, Bizcaia, Spain;Alain Vergnenegre, Service de Pathologie Respiratoire et d’Allergologie; Jose MiguelSanchez, Hospital La Princesa, Madrid, Spain; Mariano Provencio, Hospital Puerta deHierro, Madrid, Spain; Filippo de Marinis, Divisione di Oncologica Toracica, IstitutoEuropeo di Oncologia, Milano, Italy; Enric Carcereny, Institut Català d’Oncologia,Hospital Germans Trias i Pujol, Badalona, Spain; Noemi Reguart, Hospital Clínic,Barcelona, Spain; Charo Garcia Campelo, Hospital A Coruña, A Coruña, Spain; EricSantoni-Rugiu, Rigshospitalet, Copenhagen, Denmark. Finally, we thank M.K. Occhipintifor editorial assistance. The work in Dr. Ditzel’s laboratory was supported by the

Danish Cancer Society, and Sino-Danish Centre for Education and Research, NationalExperimental Therapy Partnership (NEXT) Bioinformatics financed by Innovation FundDenmark, and Academy of Geriatric Cancer Research (AgeCare). The work inDr. Rosell’s laboratory was supported by grants from the La Caixa Foundation and RedTematica de Investigacion Cooperativa en Cancer (RTICC; grant RD12/0036/ 0072).Dr. Larsen was supported by the Lundbeck Foundation (Junior Group LeaderFellowship) and a grant from the VILLUM Foundation to the VILLUM Center forBioanalytical Sciences. Dr. Bivona was supported by grants from NIH/NCIR01CA169338 and Pew and Stewart Foundations. We also thank the animal core facilityat University of Southern Denmark for maintenance and care of the mice in this study.

Author contributionsK.J., M.A.M., H.J.D. and R.R. designed the study. K.J., J.B.-A., A.G.-C., M.G., C.C.-S.,J.C.-S., M.A.M. and M.R.L. contributed with in vitro data. K.J. and M.H.P. performedthe animal study. N.K. and C.T. contributed with data on clinical cohorts. U.L., E.F. andS.V. contributed with clinical samples. A.D. contributed with statistical analysis. K.J.,N.K., M.A.M. and H.J.D. wrote the manuscript. T.G.B. revised the manuscript. R.R.,M.A.M. and H.J.D. supervised the project and revised the manuscript.

Additional informationSupplementary Information accompanies this paper at doi:10.1038/s41467-017-00450-6.

Competing interests: The authors declare no competing financial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.

Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing,

adaptation, distribution and reproduction in any medium or format, as long as you giveappropriate credit to the original author(s) and the source, provide a link to the CreativeCommons license, and indicate if changes were made. The images or other third partymaterial in this article are included in the article’s Creative Commons license, unlessindicated otherwise in a credit line to the material. If material is not included in thearticle’s Creative Commons license and your intended use is not permitted by statutoryregulation or exceeds the permitted use, you will need to obtain permission directly fromthe copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

© The Author(s) 2017

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00450-6

14 NATURE COMMUNICATIONS |8: 410 |DOI: 10.1038/s41467-017-00450-6 |www.nature.com/naturecommunications